Decentralized Adaptive Critic Control Design for a Class of Disturbed Nonlinear Interconnected Systems

Abstract

In this paper, we establish a neural-network-based adaptive dynamic programming (ADP) method to address the decentralized control problem for nonlinear continuous-time (CT) interconnected systems with external disturbances. First, the optimal control schemes of the isolated subsystems are constructed, which are based on the appropriate cost functions. Besides, it is accessible to establish the effective decentralized control strategy of the large-scale systems by adding suitable feedback gains. Then, by taking into account the disturbance rejection issue, the Hamilton-Jacobi-Isaacs (HJI) equation can be derived. To solve the HJI equation, a series of critic neural networks are developed to obtain the cost functions and the optimal control strategies approximately. Additionally, a novel weight updating rule is proposed during the neural critic learning process. Finally, the simulation results are presented to display the decentralized control performance by adopting the neural-network-based ADP method.